PSI - Issue 62

Daniela Fusco et al. / Procedia Structural Integrity 62 (2024) 895–902 Fusco et al./ Structural Integrity Procedia 00 (2019) 000 – 000

901

7

(a)

(b)

Fig. 4. Dynamic response prediction under white noise excitation through NAR model in training phase (U1): comparison between numerical response (target) and network prediction (output) of displacement time series (a) and frequency content (b).

(a)

(b)

(c)

Fig. 5. Time series prediction of displacements and prediction error of the NAR model trained in U1 conditions and tested on response in U2 (a), D6 (b) and P1 (c) scenarios under white noise excitation.

(a)

(b)

Fig. 6. NRMSE Variation (a) and Frequency Variation (b) for the considered scenarios.

Through the simulation of the dynamic response under withe noise excitation, only a neural network model related to the healthy state of the structure has been obtained and its failure in predicting the structural response of the damaged beam has been considered as damage indicator. The prediction error of such network turned out to be a suitable measure for the definition of a damage indicator able to detect the presence of the concrete cracks and reinforcement yielding. Finally, the fiber beam model and the refined constitutive law considered allowed to perform fast and reliable analyses and to accurately define the threshold level for the outlier analysis, which is a critical aspect in the unsupervised data-driven methods adopted for damage detection tasks. Acknowledgements This study was carried out within the MOST – Sustainable Mobility National Research Center (Spoke 7, WP4) and received funding from the European Union Next-GenerationEU (PNRR – MISSIONE 4 COMPONENTE 2,